Anda di halaman 1dari 57

Module F12MS3: Oscillations and Waves

Bernd Schroers 2007/08

This course begins with the mathematical description of simple oscillating systems such as a mass on a spring or a simple pendulum. We study the behaviour of such systems with and without damping, and when subjected to a periodic external force. We then study coupled oscillators: several masses connected by springs, or beads on an elastic string. The equations of motion for these systems have some particularly simple solutions, called normal modes, which we describe explicitly and in detail. By smearing out the masses of the beads on a string we construct a mathematical model for a continuous elastic medium: the elastic string. We show that the transverse displacement of such a string obeys a partial dierential equations, called the wave equation, and nd solutions which describe standing waves. The standing waves are the continuous analogue of the normal modes of the coupled oscillators. The idea that a general conguration of the string is a sum of standing waves leads into the mathematical theory of Fourier series: the decomposition of arbitrary functions in terms of cosine and sine functions. Finally we study travelling wave solutions of the wave equation. A note on units: We will be using the SI system. Length is measured in metres (m), time in seconds (s or sec), mass in kilogrammes (kg) and force in Newtons (N). The Newton is a derived unit and can be expressed in terms of m, s and kg: 1 N=kg m/s2 . Note that this is consistent with Newtons second law Force=mass acceleration. In these units the gravitational acceleration on earth is g = 9.8 m/s2 =9.8N/kg.

Table of Contents
1 Simple harmonic motion 1.1 1.2 1.3 1.4 1.5 Hookes law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Properties of simple harmonic motion . . . . . . . . . . . . . . . . . . . . . Energy in simple harmonic motion . . . . . . . . . . . . . . . . . . . . . . The vertical spring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The simple pendulum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 3 4 4 7 8 10

2 Revision interlude 2.1 2.2

Complex numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Second order dierential equations with constant coecients . . . . . . . . 12 2.2.1 2.2.2 Homogeneous equations with constant coecients . . . . . . . . . . 12 Inhomogeneous equations . . . . . . . . . . . . . . . . . . . . . . . 14 1

3 Damped and forced oscillations 3.1 3.2 3.3 3.4 3.5

18

The oscillating spring revisited . . . . . . . . . . . . . . . . . . . . . . . . . 18 Unforced oscillations with damping . . . . . . . . . . . . . . . . . . . . . . 20 Forced oscillations with damping . . . . . . . . . . . . . . . . . . . . . . . 22 Forced oscillations without damping . . . . . . . . . . . . . . . . . . . . . . 25 Energy in damped and driven oscillations . . . . . . . . . . . . . . . . . . . 26 28

4 Coupled oscillators 4.1 4.2 4.3

Coupled springs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 N coupled oscillators: beads on a string . . . . . . . . . . . . . . . . . . . . 31 The eigenvector method for nding normal modes . . . . . . . . . . . . . . 34 39

5 Waves 5.1 5.2 5.3 5.4 5.5

The wave equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Standing waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Fourier series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 Using Fourier series for solving the wave equation . . . . . . . . . . . . . . 50 Travelling waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

1
1.1

Simple harmonic motion


Hookes law

Consider a spring attached to a wall and lying horizontally on a smooth surface. It is found empirically that the force required to stretch the spring elastically is proportional to the stretching. This result is named after the English physicist Robert Hooke (1635-1703): Physical Law 1.1.1 (Hookes law) The force FH required to stretch the spring by an amount x is given by FH = kx, (1.1)

where k is a constant of proportionality which is characteristic of the spring and called the spring constant. Suppose we now attach a trolley of mass m to the free end of the spring. When the trolley is displaced from the equilibrium position by an amount x the spring exerts a force on the trolley which, by Newtons third law, is equal and opposite to the force FH exerted by the trolley on the spring. It is therefore given by F = kx. (1.2)

and sometimes called the elastic restoring force. According to Newtons second law the motion of the trolley will be such that its acceleration x satises mx = kx. This is the equation of simple harmonic motion (SHM). With the denition 2 = the equation becomes x = 2 x (1.5) k m (1.4) (1.3)

It is easy to check that x(t) = cos(t) and x(t) = sin(t) solve (1.6). Further solutions are A cos(t) for any constant A, but also the sum x(t) = cos(t) + sin(t). More generally, we have the following Theorem 1.1.2 (Principle of superposition) If x1 (t) and x2 (t) are solutions of (1.5), then so is x(t) = Ax1 (t) + Bx2 (t), where A and B are constants. To prove this theorem, compute x = Ax 1 + B x 2 . But since x 1 = 2 x1 and x 2 = 2 x2 , we have x = A 2 x1 B 2 x2 = 2 x as required. 3

The principle of superposition allows one to generate new solutions from two given solutions. If certain conditions on the dierential equation and the two given solutions are satised one can show that all solutions are obtained from the superposition principle. This is the case for the solutions x1 (t) = cos(t) and x2 (t) = sin(t) of (1.5). The general solution is therefore x(t) = A cos(t) + B sin(t). General solutions of second order dierential equations depend on two constants. 1.2 Properties of simple harmonic motion (1.6)

We can write the solution (1.6) also in the form x(t) = R cos(t ), (1.7)

where the angle lies in the interval (, ]. Using the trigonometric identity cos( ) = cos cos +sin sin , and comparing with (1.6) we deduce the following relation between the constants A, B on one hand and R, on the other: R cos = A, and R sin = B. (1.8)

These can be inverted to give R = A2 + B 2 . We also deduce tan = B/A, but note that, since tan( + ) = tan(), this only determines up to multiples of . To get the correct value of you need to refer back to (1.8). R is the furthest distance the trolley travels from the equilibrium position during the motion and is called the amplitude of the oscillation. Since cos is a periodic function with period 2 , the motion repeats itself after a time T which is such that T = 2 , i.e. T = 2 . (1.9)

This is called the period of the motion. The inverse = 1/T is the frequency and = 2 is called the angular frequency. The angular frequency = k/m is sometimes called the characteristic frequency. Example 1.2.1 A trolley of mass m = 1 kg is attached to a spring with spring constant k = 64N/m. The trolley is pulled 1/4 m to the right of the equilibrium position and released from rest. Find its subsequent motion. What is its amplitude and period?
1 x(t) = 4 cos 8t. Amplitude R = 1/4 m, period T = /4 seconds.

1.3

Energy in simple harmonic motion

Kinetic energy is a measure of the energy in the motion of a particle and was discussed in the module Mathematics of motion. We recall the denition: 4

Denition 1.3.1 (Kinetic energy) The kinetic energy K of a particle of mass m moving with velocity x is 1 K = mx 2. 2 (1.10)

The potential energy of a particle attached to a spring is a measure of the energy stored in the spring-particle system when the spring is stretched or compressed. The precise denition is motivated by the requirement that the sum of kinetic energy and potential energy should be conserved. Denition 1.3.2 (Potential energy) The potential energy of simple harmonic motion is 1 V = kx2 . 2 With these denitions we have Theorem 1.3.3 Energy conservation The total energy E = K + V is conserved during simple harmonic motion. Proof: Dierentiating, using the product rule and chain rule, we nd dE = mx x + kxx =x (mx + kx) = 0, dt by virtue of the equations of motion (1.3). Both the kinetic and potential energy change during the motion, but a good measure of how the total energy is divided into kinetic and potential energy is given by the average kinetic and potential energy. Denition 1.3.4 The average kinetic energy is Kav = and the average potential energy is Vav 1 = T
T 0

(1.11)

(1.12)

1 T

T 0

1 mx 2 dt 2

(1.13)

1 2 kx dt. 2

(1.14)

Example 1.3.5 Compute the kinetic and potential energy, the total energy and the average kinetic and potential energy for the motion of the trolley in example 1.2.1

1 cos 8t, For the potential energy, measured in Joule, we nd, with x(t) = 4

V =

1 1 64 ( cos 8t)2 = 2 cos2 8t 2 4

(1.15)

and for the kinetic energy, also measured in Joule, we use x (t) = 2 sin 8t and nd K= 1 (2 sin 8t)2 = 2 sin2 8t 2 (1.16)

Hence the total energy in Joule is E = 2(cos2 8t + sin2 8t) = 2. (1.17)

In order to compute the average kinetic and potential energy we use the trigonometric identities 1 1 sin2 z = (1 cos(2z )), cos2 z = (1 + cos(2z )). (1.18) 2 2 Combine them with the fact that the denite integrals of both cos(nz ) and sin(nz ) from 0 to 2 vanish for any integer n
2 2

sin(nz ) dz =
0 0

cos(nz ) dz = 0.

(1.19)

to conclude
2 2

sin2 z dz =
0 0

1 1 dz 2 2

cos(2z ) dz =
0

(1.20)

and similarly
2

cos2 z dz = .
0

(1.21)

Then, with T = /4 and z = 8t Kav 8 =


4

sin2 (8t) dt =

sin2 z dz = 1
0

(1.22)

By a very similar calculation Vav = 1. (1.23)

In the example we saw that both the kinetic and the potential energy oscillate in such a way that one is zero when the other is maximal: the total energy changes back and forth between kinetic and potential energy. The sum of kinetic and potential energy is constant and on average there is an equal amount of both in every cycle of the oscillation. The conclusions generalise from the example to general simple harmonic motion. Its worth recording one general result here, which we used in the example and which we will need again. 6

Lemma 1.3.6 Consider an arbitrary angular frequency and the associated period T = 2 . Then 1 T
T 0

1 sin (t) dt = T
2

sin2 (t) dt =
0

1 2

(1.24)

To prove this, let z = t, so that 1 T and 1 T


T T

sin2 (t) dt =
0

1 2 1 2

sin2 z dz
0

(1.25)

cos2 (t) dt =
0

cos2 z dz
0

(1.26)

Now use the results (1.21) and (1.19) to obtain the claim. 1.4 The vertical spring

Suppose a spring his hanging vertically as shown in in Fig. 1.1 and an object of mass m is attached to the free end of the spring, thus exerting a gravitational force gm. According to Hookes law (1.1) the spring will stretch by an amount l satisfying mg = kl. (1.27)

Figure 1.1: The vertical spring We are interested in the motion of the object when we stretch the spring by an additional amount x0 and release it, possibly with some initial velocity v0 . Ignoring friction, the forced acting on the spring are the gravitational force mg (downwards) and the elastic restoring force k (l + x). Hence, by Newtons second law, the equation of motion is m d2 x = mg k (l + x) = kx dt2 (1.28)

where we used (1.27). Thus we nd that the eect of gravity disappears from the equation, and the displacement x from the equilibrium obeys the same equation as the horizontal spring! 7

49 m. If the object is Example 1.4.1 Suppose an object of 1 kg mass stretches a spring 20 1 given an initial speed 5 m s downwards at the equilibrium position, nd the subsequent motion. Neglect air resistance.

From the data given and equation (1.27) we compute the spring constant k= mg 20 = 9.8 N/m = 4N/m. l 49

Since m = 1kg we nd = 2s1 . The general solution is thus x(t) = A cos(2t) + B sin(2t), (1.29)

where it is understood that t gives the time measured in seconds and x the distance from equilibrium measured in metres. From the initial condition we deduce A = 0 and B = 5/2 and thus x(t) = 5 sin(2t). 2 (1.30)

1.5

The simple pendulum

The simple pendulum is a bob of mass m suspended from a xed point O by a light, inextensible rod (or string) of length l. The rod is hinged and allowed to move in one plane only. The equation of motion for the simple pendulum can be derived in two ways, either using the vector description of two-dimensional motion developed in the module Mathematics of motion or using rotational dynamics of rigid bodies. Since we have not covered the latter I give a derivation based on the rst method. This is a little intricate and you are welcome to skip to the result (1.43) if you are willing to take it on trust. First we set up some notation to describe the position of bob, see also Fig. (1.2). In terms of the coordinate axis i and j and the angle shown in the gure the position vector of the bob relative to O is r(t) = l(sin i cos j ). (1.31)

When the bob moves, the angle changes with time but l is constant since the rod is supposed to be inextensible. Thus the velocity is = l (cos i + sin j ) r and the acceleration is = l (cos i + sin j ) l 2 (sin i cos j ). r With the abbreviation n = sin i cos j 8 (1.34) (1.33) (1.32)

j
O

t n

Figure 1.2: The simple pendulum for the normalised vector in the direction of the rod and t = cos i + sin j for the normalised tangential vector we can write the acceleration as = l t l 2 n r (1.36) (1.35)

Neglecting air resistance, there are two forces on the bob, namely its weight and the tension in the rod. The weight W has magnitude mg and points in the j direction; the tension always points in opposite direction to r and we denote its magnitude by . The total force on the bob is therefore F = (sin i cos j ) mg j. Expressing j in terms of n and t as j = sin t cos n we write the force as F = n mg sin t + mg cos n. 9 (1.38) (1.37)

Thus decomposing Newtons second law = F mr into its t and n component we nd = mg sin ml and 2 = + mg cos . m (1.41) (1.40) (1.39)

Of the two equations we found, the rst (1.40) is the equation of motion for the bob; the equation (1.41) merely determines the tension in the rod resulting from the bobs motion. The equation (1.40) is dicult to solve in general, but if we restrict ourselves to small oscillations we can approximate sin and the equation becomes = g . l This is the equation for simple harmonic motion with angular frequency = so that the period of the oscillation is T = 2 l . g (1.45) g l (1.44) (1.43) (1.42)

Thus a long pendulum swings more slowly than a short one, and a pendulum on the moon more slowly than the same pendulum on earth. Moreover, the period is independent of the mass of the bob! This should be contrasted with the situation for the mass on a spring.

2
2.1

Revision interlude
Complex numbers

Complex numbers are pairs of real numbers x, y R and written as z = x + iy (2.1)

where i is called the imaginary unit. The real numbers x and y are called the real and imaginary part of z , written as x = Re(z ), y = Im(z ). 10 (2.2)

The set of all complex numbers is denoted C. Addition of complex numbers is dened component-wise. If z1 = x1 + iy1 and z2 = x2 + iy2 then z1 + z2 = x1 + x2 + i(y1 + y2 ). The multiplication rule is determined once we x i2 = 1. Then z1 z2 = x1 x2 y1 y2 + i(x1 y2 + x2 y1 ). The complex conjugate z of the complex number z is dened as z = x iy and satises zz = x2 + y 2 . It follows that the inverse of z is 1 z x iy . = = 2 z zz x + y2 (2.8) (2.7) (2.6) (2.5) (2.4) (2.3)

Im(z)

z R x Re(z)

Figure 2.1: The Argand diagram Complex numbers can be depicted as vectors in the two-dimensional plane, called the Argand diagram, where the real part is plotted along the x-axis and the imaginary part along the y -axis. The length R of the vector representing the complex number z is the called modulus and the angle (measured in radians) it makes with the x-axis is called the argument of z . The angle is only dened up to multiples of 2 , but we adopt the convention that lies in the interval (, ] (often called the principal value). Thus the modulus is given by R = x2 + y 2 = z z (2.9) 11

and the argument satises tan = From the gure we nd x = R cos , or z = R(cos + i sin ). Using the important Euler relation ei = cos + i sin we have the modulus-argument representation of z : z = Rei . (2.14) (2.13) (2.12) y = R sin (2.11) y . x (2.10)

The modulus of z is also sometimes written as |z |, so |z | = R when z is given by (2.14). The multiplication of complex numbers is particularly simple in the modulus-argument form. If z1 = R1 ei1 and z2 = R2 ei2 then z1 z2 = R1 R2 ei(1 +2 ) . The inverse of z in (2.14) is 1 1 = ei z R and its complex conjugate is z = Rei . (2.17) (2.16) (2.15)

To end, we note an important consequence of the relation (2.13) and its complex conjugate ei = cos i sin . Adding and subtracting (2.13) and (2.18) we nd cos = 2.2 2.2.1 1 i e + ei 2 and sin = 1 i e ei . 2i (2.19) (2.18)

Second order dierential equations with constant coecients Homogeneous equations with constant coecients

Consider d2 x dx + a + a0 x = 0, 1 dt2 dt 12 (2.20)

where a1 and a0 are real constants. We try solutions of the form x(t) = et . Inserting (2.21) into (2.20) leads to (2 + a1 + a0 )et = 0. (2.22) (2.21)

Since et is never zero, we deduce that (2.21) is a solution if satises the equation 2 + a1 + a0 = 0. (2.23)

This is called the characteristic equation of the dierential equation (2.20). Its roots are 1 = a1 + a2 1 4a0 2 and 2 = a1 a2 1 4a0 2 (2.24)

Inserting the roots into (2.21) we thus obtain solutions to the dierential equation (2.20. The nature of the solution depends on the roots. (i) 1 = 2 real. This is the easiest case: x1 (t) = e1 t and x2 (t) = e2 t are independent solutions: they form a fundamental set of solutions. The general solution is a linear combination of x1 and x2 : x(t) = Ae1 t + Be2 t , with A and B constants. (ii) 1 = 2 real. In that case we obtain only one solution from (2.21), namely x1 (t) = e1 t . A second independent solution is given by x2 (t) = te1 t , and the general solution is of the form x(t) = (A + Bt)e1 t (2.26) (2.25)

1 , so if 1 = p + iq then 2 = p iq . The (iii) 1 , 2 complex. In that case 2 = functions e1 t and e2 t are independent solutions of (2.20) but are complex. To obtain real solutions we take the linear combinations 1 x1 (t) = (e(p+iq)t + e(piq)t ) = ept cos(qt) 2 1 (p+iq)t x2 (t) = (e e(piq)t ) = ept sin(qt) 2i and obtain a real fundamental set. The general solution is of the form x(t) = ept (A cos(qt) + B sin(qt)). (2.28)

(2.27)

13

roots

fundamental set of solutions

1 = 2 real 1 = 2 real 1 = p + iq, 2 = p iq

u1 (t) = e1 t , u1 (t) = e1 t , u1 (t) = ept cos(qt),

u2 (t) = e2 t u2 (t) = te1 t u2 (t) = ept sin(qt)

Table 1: fundamental sets for constant coecient equations 2.2.2 Inhomogeneous equations

Consider now dx d2 x + a1 + a0 x = f, 2 dt dt where f is some function of t. In order to nd all solutions of (2.29) we require 1. A fundamental set {x1 , x2 } of solutions of 2. A particular solution xp of (2.29). The general solution is then given by x = Ax1 + Bx2 + xp , where A and B are real constants. If the function f in (2.29) is a polynomial, exp, sin or cos, then one can nd a particular solution of a similar form via the method undetermined coecients. The following table gives recipes involving unknown coecients which one can determine by substituting into the equation. There is no deep reason for these recipes other than that they work. In the table I have abbreviated homogeneous equation by HE. (2.30) dx d2 x + a1 + a0 x = 0. 2 dt dt (2.29)

14

f (t)

particular solution

b0 + b1 t + ...bn tn et

c0 + c1 t + ...cn tn et is not a solution of HE try cet et is a solution of HE try ctet t e and tet solutions of HE try ct2 et eit not solution of the HE try Ceit , C C, and take real part eit solution of the HE try Cteit , C C, and take real part Find particular solution for equation with f (t) = eit and take real part

eit

cos(t)

Table 2

To illustrate the recipes given in the table, we consider some examples, beginning with the polynomial case. d2 x + x = t2 . dt2 We try xp (t) = b0 + b1 t + b2 t2 , and nd by inserting into (2.31) (2b2 + b0 ) + b1 t + b2 t2 = t2 Comparing coecients yields b2 = 1, so that xp (t) = t2 2. Continuing with the exponential case, consider dx d2 x + 3 + 2x = e3t 2 dt dt 15 (2.36) (2.35) b1 = 0, b0 = 2 (2.34) (2.33) (2.32) (2.31)

The roots of the characteristic polynomial are 1 and 2. Thus the right hand side is not a solution of the homogeneous equation, and we try xp (t) = ce3t (2.37)

and deduce from inserting into (2.36) that xp is a solution provided we choose c = 1/20. Now change the right hand side of (2.36) to a solution of the homogeneous equation, say dx d2 x + 3 + 2x = et . 2 dt dt Now we try xp (t) = ctet . (2.39) (2.38)

After slightly tedious dierentiations, you should nd that this is indeed a solution provided we pick c = 1. Finally we turn to the oscillatory case f (t) = cos(t). It is particularly important for applications and we will study it in detail in Sect. 3. According to method summarised given in the last row of table 2 we should solve the equation with cos(t) replaced by eit , and take the real part at the end. The method can be justed as follows. Suppose x(t) is a solution of x + a1 x + a0 x = eit , with a1 , a0 and all real. Then, taking real parts, Re ( x + a1 x + a0 x) = Re eit = cos(t) (2.41) (2.40)

so, with xp = Re(x), we have Re( x) = x p , Re(a1 x ) = a1 x p and Re(a0 x) = a0 xp . Hence x p + a1 x p + a0 xp = cos(t) To illustrated the method, we study one example here: d2 x dx + 3 + 2x = cos(2t). 2 dt dt We consider the associated complex equation d2 x dx + 3 + 2x = e2it . 2 dt dt (2.44) (2.43) (2.42)

Our strategy is to solve this equation, and take the real part of the (complex) solution as a particular solution for (2.43). Thus we try x(t) = Ce2it with C C. Dierentiating and inserting into (2.44) we nd (4C + 6iC + 2C )e2it = e2it 16 (2.45)

so that C= Thus the particular solution is xp (t) = Re 3 1 i (cos(2t) + i sin(2t)) 20 20 1 3 = cos(2t) + sin(2t). 20 20 1 1 3 = i. 2 + 6i 20 20 (2.46)

(2.47)

17

3
3.1

Damped and forced oscillations


The oscillating spring revisited

Consider a spring hanging vertically as in 1.4, with an object of mass m attached to it. We are now going to consider the more complicated situation where the object is acted on by an additional force f (t), and we are going to take friction into account. The forces acting on the object are 1. The downward gravitational force: mg . 2. The elastic restoring force k (l + x). 3. Air resistance. This is an example of a damping force which is proportional to the . velocity but acts in the opposite direction: r dx dt 4. Any other force exerted on the object, denoted f (t). Note that the damping coecient r has units N s/m and the spring constant k has units N/m. According to Newtons second law, the motion of the object is thus governed by the equation m dx d2 x = mg k (l + x) r + f (t). 2 dt dt (3.1)

Re-arranging the terms, and using (1.27) we thus arrive at the linear second-order inhomogeneous ODE m d2 x dx + r + kx = f (t). dt2 dt (3.2)

This is the sort of equation which we have learnt to solve in the previous subsection. Here we will see how to interpret our solutions physically. I have summarised the physical meaning of the various parameters and functions in the following table.

18

x(t) x (t) x (t) f (t) m r k

downward displacement from equilibrium at time t velocity at time t acceleration at time t external or driving force at time t mass damping coecient spring constant Table 3

We are now going to study the equation (3.2), starting with the simplest situation and building up towards the general form (3.2). First consider the case where the damping constant r is zero, and no additional force f acts on the mass. The equation (3.2) thus becomes d2 x k + x = 0. dt2 m (3.3)

This is the equation of simple harmonic motion we studied at the beginning of the course. Lets derive the solution using the general technique developed in subsection 2.2. The characteristic equation is 2 + With the abbreviation 0 = a fundamental set of solutions is given by x1 (t) = cos(0 t), The general solution is therefore x(t) = A cos(0 t) + B sin(0 t) in agreement with (1.6). (3.7) x2 (t) = sin(0 t). (3.6) k m (3.5) k = 0. m (3.4)

19

3.2

Unforced oscillations with damping

There is no external force, but the damping coecient r is not zero. The equation is m or dx d2 x 2 + + 0 x=0 2 dt dt with the abbreviations = r m (3.10) (3.9) d2 x dx +r + kx = 0 2 dt dt (3.8)

and 0 as in (3.5). The characteristic equation is


2 2 + + 0 =0

(3.11)

with roots 1 = + 2 2 2 0 4 and 2 = 2 2 2 0 4 (3.12)

Provided the roots are dierent, a fundamental set is therefore given by u1 (t) = e1 t , u2 (t) = e2 t . (3.13)

Both physically and mathematically the discussion of the general solution is best organised 2 according to the sign of 2 40 .
2 (i) Underdamped case: 2 < 40

We introduce the abbreviation =


2 0

2 4

(3.14)

so that 1 = /2 + i and 2 = /2 i . Comparing with table 1 in section 2.5 we deduce that a real fundamental set is given by x1 (t) = e 2 t cos(t) The general solution is therefore given by x(t) = e 2 t (A cos(t) + B sin(t)) = R e 2 t cos(t ) with R and dened as after Eq. (1.6). The solution still oscillates and has innitely many zeroes, but the amplitude of the oscillation decreases exponentially with time. For t all solutions tend to zero: the oscillation dies down. 20

x2 (t) = e 2 t sin(t).

(3.15)

(3.16)

1 0.8 0.6 0.4 0.2 0 0.2 0.4 1 2 3 4 t 5 6 7 8

Figure 3.1: Plot of the underdamped solution (3.16) for A = 1, B = 0.25, = 4, = 2


1

0.8

0.6

0.4

0.2

4 t

Figure 3.2: The overdamped solution (3.17) for A = 1.5, B = 0.5, 1 = 0.5, 2 = 1.5
2 (ii) Overdamped case: 2 > 40

In this case both roots 1 and 2 (3.12) are real and negative. The general solution is x(t) = Ae1 t + Be2 t . (3.17)

Solutions are zero for at most one value of t (provided A and B do not both vanish) and tend to 0 for t .
2 (iii) Critically damped case: 2 = 40

In this case 1 = 2 = /2 and (compare sect. 2.4.3) the general solution is x(t) = (A + Bt)e 2 t

(3.18)

Again solutions are zero for at most one value of t (provided A and B do not both vanish) and tend to 0 for t . The terminology underdamped, overdamped and critically damped has its origin in engineering applications. The theory developed here applies, for example, to the springs that provide the damping in cars. When perturbed from equilibrium, underdamped springs return to the equilibrium position quickly but overshoot. Overdamped springs take a long time to return to equilibrium. In the critically damped case the 21

0.8

0.6

0.4

0.2

4 t

10

Figure 3.3: Plot of the critically damped solution (3.18) for A = B = 1, = 2 spring returns to the equilibrium position very quickly but avoids overshooting. Thus the critically damped case provides the most ecient damping. Example 3.2.1 An object of mass m = 2 kg is attached to a spring with spring constant k = 4 N m1 and immersed in a viscous liquid with damping constant r = 6 N m1 s. At time t = the object is raised 1 m and given an initial downward velocity of 3 m s1 . Find the subsequent motion of the object. The equation of motion is x + 3x + 2x = 0. (3.19)

The characteristic equation has the two real roots 1 = 2 and 2 = 1, so the general solution is x(t) = Aet + Be2t (3.20)

Since x(0) = A + B = 1 and x (0) = A 2B = 3 we deduce A = 1 and B = 2. The solution x(t) = et 2e2t passes through the equilibrium x = 0 once at t = ln 2 s. 3.3 Forced oscillations with damping (3.21)

This is the most general case, with all terms in eq. (3.2) playing a role. If the external force f grows indenitely, it is clear that the spring will eventually break. Remarkably this can also happen when f is a periodic function which averages to zero. Since this case is particularly important in applications, we focus on it here. Suppose therefore that f (t) = f0 cos(t), 22 (3.22)

where f0 is some constant (in units of Newtons). Then the equation of motion is m d2 x dx +r + kx = f0 cos(t). 2 dt dt (3.23)

After dividing by m the equation takes the form f0 dx d2 x 2 + + 0 x= cos(t). 2 dt dt m (3.24)

The quickest way to solve this is to use complex numbers. Since cos t is the real part of exp(it) we solve f0 it d2 x dx 2 + + x = e . 0 dt2 dt m (3.25)

rst and then take the real part of the solution we obtain. This turns out to be an ecient method. Suppose rst that i is not a solution of the characteristic equation. Then try x(t) = C exp(it). Inserting into (3.25) yields
2 Ceit ( 2 + i + 0 )=

f0 it e . m

(3.26)

Dividing by exp(it) and solving for C we nd C= where tan = 2 2 0 (3.28) (f0 /m) = 2 2 + i + 0 (f0 /m)ei
2 (0 2 )2 + 2 2

(3.27)

Taking the real part of x(t) = C exp(it) we therefore nd the solution xp (t) = The function R( ) = (f0 /m)
2 (0 2 )2 + 2 2

(f0 /m)
2 (0

2 )2 + 2 2

cos(t ).

(3.29)

(3.30)

gives the -dependent amplitude of the forced motion. For later use we write it as R( ) = f0 m0 1
0

(3.31) +
2 2 0

The general solution of (3.24) is a linear combination of the particular solution (3.29) and the fundamental solutions of the free damped system discussed in sect. 2.6.2. Let us for 23

2 deniteness consider the underdamped case 2 < 40 . Then the relevant fundamental set is (3.15) and the general solution of the inhomogeneous equation (3.24) is

x(t) = e 2 t (A cos(t) + B sin(t)) + transient solution

xp (t)

(3.32)

steady state solution

with and as dened in (3.14). The rst part is called the transient solution because it tends to zero for t . After a long time the solution is dominated by the steady state solution. Note that this has the same frequency as the driving term (3.22). Also note that the amplitude of the steady state solution depends on the frequency and is very large when = 0 . This phenomenon is called resonance. Resonance occurs when the driving frequency is equal to the characteristic frequency of the spring. Depending on the value of r, m and the amplitude of the steady state solution at resonance could be much bigger than the amplitude of the driving force. Finally note that at resonance tan = so that = /2. Thus the steady solution lags behind the driving force by /2 at the resonance frequency. Resonance manifests itself more dramatically when the damping is small or zero, as we shall see in the next section.
1

0.8

0.6

0.4

0.2

4 omega

Figure 3.4: The amplitude (3.30) for f0 = 2, 0 = 2, = 1 The amplitude of displacement is not maximal at resonance. We now show that the amplitude of the velocity is maximal there. Dierentiating (3.29) we have x p (r) = R( ) sin(t ). This is an oscillation with amplitude V ( ) = R( ) = f0 m0 1
0 2

(3.33)

(3.34) +
2
2 0

where we used the expression (3.31). Written in this way it is obvious the denominator is minimal as a function of when = 0 . Hence the velocity amplitude V is maximal at resonance. 24

1.5

0.5

4 omega

Figure 3.5: The velocity amplitude (3.34) for f0 = 2, 0 = 2, = 1 3.4 Forced oscillations without damping

In the absence of damping, the equation (3.23) simplies to m d2 x + kx = f0 cos(t). dt2 (3.35)

This is a special case of the discussion of oscillatory driving terms in sect. 2.5.1. First consider the case = 0 , i.e. i is not a root of the characteristic equation. Then we have the particular solution xp (t) = and therefore the general solution is x(t) = A cos(0 t) + B sin(0 t) + f0 /m cos(t) 2 (0 2) (3.37) (f0 /m) cos(t) 2 (0 2) (3.36)

which is a fairly complicated superposition of oscillations of dierent frequencies. It remains bounded as t . Next consider the case of resonance. The driving frequency equals the springs characteristic frequency, i.e. = 0 . Again we think of cos(t) as the real part of exp(it) and study d2 x + 2 x = eit dt2 (3.39) (3.38)

Since i is a solution of the characteristic equation, we try x(t) = Ct exp(it) in accordance with Table 2. We nd 2iCeit = eit 25 (3.40)

or C= 1 . 2i
t 2

(3.41) (i cos(t) + sin(t)) we obtain the

Taking the real part of x(t) = Ct exp(it) = particular solution xp (t) = The general solution is therefore

t sin(t) 2 f0 t sin(t). 2m

(3.42)

x(t) = A cos(t) + B sin(t) +

(3.43)

The remarkable property of the particular solution is that its amplitude grows linearly with time and becomes innite as t . In real life ever increasing oscillations mean that the oscillating system (be it a spring or a more complicated object - such as a building) will break. This can lead to very dramatic manifestations of resonance, as for example in the collapse of the Tacoma Narrows bridge in the USA in 1940. 3.5 Energy in damped and driven oscillations

In the presence of damping and an external force, the energy of an oscillating spring is no longer conserved. Instead we nd, with E = 1 mx 2 + 1 kx2 dened as in Sect. 1.3, 2 2 dE = (m x + kx)x = rx 2 + xf, dt (3.44)

where we used the equation of motion (3.23) and f (t) = f0 cos(t). The terms on the right hand describe how the energy of the spring changes. The rst term is always negative and describes the energy lost per unit time due to friction. The second term describes the rate at which energy is put into or taken out of (depending on the sign) the spring by the external force. Rate of change of energy is called power in physics, so we dene the dissipated power as Pdis = rx 2 and the power input as P = fx (3.46) (3.45)

Lets compute the power input for the forced and damped spring. For the steady state solution (3.29) we have x b = R( ) sin(t ) and hence P (t) = f0 R( ) cos(t) sin(t ) = f0 R( ) (cos(t) sin(t) cos cos(t) cos(t) sin ) 26 (3.48) (3.47)

1 sin(2t) and Using cos(t) sin(t) = 2

1 T as well as 1 T

sin(2t)dt = 0
0

(3.49)

cos2 (t)dt =
0

1 2

(3.50)

(see (1.24)) we nd that the average power input over one cycle is = 1 P T With the expression R( ) = f0 m0 1
0 T 0

1 P (t)dt = f0 R( ) sin 2

(3.51)

(3.52) +
2 2 0

(see (3.31)) for R and sin = = 0


2 (0

2 )2 + 2 2 1
0

(3.53)

2
2 0

we have
2 = f0 P 2 2m0

1
0

(3.54) +
2 2 0

, like the velocity amplitude V ( ), is maximal when Comparing with (3.34) we see that P = 0 .

27

4
4.1

Coupled oscillators
Coupled springs
x1 k K x2 k

Figure 4.1: Coupled springs Consider two particles, both of mass m, connected by a spring of spring constant K as shown in the gure. The particles lie on a horizontal plane and are connected to walls by springs with spring constant k . Their motion is restricted to a line containing all three springs. The force on each particle is obtained by applying Hookes law to the two springs connected to the particle. Counting displacement of the particles to the right as positive, the force exerted by the spring to the left of particle 1 on particle 1 is kx1 . The middle spring is stretched by an amount (x2 x1 ) when the the particles are in the position shown in the gure. It will therefore exert a force K (x1 x2 ) on particle 1 and a force K (x2 x1 ) on particle 2. Finally the spring on the right exerts a force kx2 on particle 2. Hence, by Newtons second law we have the equations of motion mx 1 = kx1 K (x1 x2 ) mx 2 = kx2 K (x2 x1 ). (4.1)

This is a system of coupled, linear second order dierential equations which looks dicult to solve. However, if we add the equations we obtain m( x1 + x 2 ) = k (x1 + x2 ) and when we subtract them we get m( x1 x 2 ) = k (x1 x2 ) 2K (x1 x2 ). Hence, with abbreviations y1 = x1 + x2 , and
2 1 =

(4.2)

(4.3)

y2 = x1 x2

(4.4)

k , m

2 2 =

k K +2 m m

(4.5)

28

we have two decoupled equations


2 y 1 = 1 y1 2 y 2 = 2 y2 .

(4.6)

Each of them is just the equation for simple harmonic motion discussed in Sect. 1. Hence the general solution is y1 (t) = A1 cos(1 t) + B1 sin(1 t) y2 (t) = A2 cos(2 t) + B2 sin(2 t) (4.7)

where A1 , B1 , A2 , B2 are arbitrary real constants. Having solved the equation (4.6) in terms of the coordinates y1 and y2 we recover the general solution x1 and x2 of (4.1) via 1 (y1 (t) + y2 (t)) 2 1 x2 (t) = (y1 (t) y2 (t)). 2 x1 (t) =

(4.8)

The coordinates y1 and y2 are called normal mode coordinates, or simply normal modes of the coupled springs. To understand their physical interpretation, consider the motion when only one of them is non-vanishing. 1. Assume that y2 = 0 (i.e. A2 = B2 = 0 in (4.7)), and consider the motion of the particles according to (4.8) when y1 = 0 . Since x1 = x2 in this case, the two particles oscillate in tandem. The spring in the middle remains unstretched since x1 x2 = 0. 2. Assume that y1 = 0 (i.e. A1 = B1 = 0 in (4.7)), and consider the motion of the particles according to (4.8) when y2 = 0 . Since x1 = x2 in this case, the two particles oscillate in opposition. The spring in the middle gets stretched and compressed in every cycle of the motion. Since 1 < 2 , the period T1 = 2/1 of the mode y1 is larger than the period T2 = 2/2 of y2 . The normal mode y1 is therefore sometimes called the slow mode, and y2 the fast mode. The following example illustrates how normal modes can be used to solve initial value problems. Example 4.1.1 Two particles of mass m = 1 kg are connected to springs as shown in Fig. (4.1). The spring constant of the outer springs is k = 25 N/m and the spring constant of the middle spring is K = 11/2 N/m . Particle 1 is initially displaced 1 m to the right and released from rest; particle 2 is initially at rest and at the equilibrium position. Find the subsequent motion of both particles.

29

We solve this problem in three steps. Step 1: Translate the initial values into intial values for the normal modes y1 and y2 . From the question we know that x1 (0) = 1, x 1 (0) = 0 and x2 (0) = 0, x 2 (0) = 0. Hence the initial values for the normal coordinates are y1 (0) = x1 (0) + x2 (0) = 1, and y2 (0) = x1 (0) x2 (0) = 1, y 2 (0) = x 1 (0) x 2 (0) = 0 y 1 (0) = x 1 (0) + x 2 (0) = 0

Step 2: Solve the initial value problem for the normal modes. With k and K given in the example, we nd from (4.5) that 1 = 5 s1 and 2 = 6 s1 . Hence the general solution of the normal mode equations (4.6) are y1 (t) = A1 cos(5t)+ B1 sin(5t) and y2 (t) = A2 cos(6t)+ B2 sin(6t). Imposing the initial conditions found in step 1 gives y1 (t) = cos(5t) and y2 (t) = cos(6t). Step3 : Translate the results of step 2 into the particle coordinates x1 and x2 . From the 1 formula (4.8) we nd x1 (t) = 1 (cos(5t) + cos(6t)) and x2 (t) = 2 (cos(5t) cos(6t)). 2 Remarks on beats It is instructive to re-write the superposition of two oscillations of dierent frequencies which appear in the nal answer for the displacements x1 and x2 in the previous question. Returning to the general situation where the normal mode frequencies are denoted 1 and 2 , consider the superposition x(t) = cos(1 t) + cos(2 t) (4.9)

and suppose that 2 1 is small compared to 1 + 2 (like in the example). Then, using the trigonometric identity 2 cos A cos B = cos(A + B ) + cos(A B ), we have x(t) = 2 cos 2 1 t cos 2 1 + 2 t . 2 (4.11) (4.10)

We interpret this formula as follows. The function A(t) = 2 cos 2 1 t 2 (4.12)

is the slowly varying amplitude of an oscillation with the rapid angular frequency (1 + 2 )/2. This is illustrated in Fig 4.2 where we plot (4.12) for the values 1 = 5 s1 and 2 = 6 s1 from the example. In physics, the periodic uctuations of the amplitude of an oscillation are called beats. The time interval between two successive maxima of the modulus of the amplitude is called the period Tbeats of the beats, and the angular frequency of the beats is dened as beats = 2 . Tbeats 30 (4.13)

6 t

10

12

Figure 4.2: Beats: plot of cos(5t) + cos(6t) against t. The dashed curve shows the slowlyvarying amplitude 2 cos(t/2) The angular frequency of beats which occur in the superposition (4.12) of is twice the angular frequency of the amplitude oscillation A(t) in (4.12) since the modulus of A(t) has half the period of A(t) . Thus beats = 2 1 . (4.14)

Note that Beats are used in tuning instruments. When two notes are slightly out of tune the dierence of their frequencies is small, leading to slow and clearly audible uctuations in the amplitude of their combined sound. The notes can be tuned by reducing the frequency of these beats, ideally to zero. 4.2 N coupled oscillators: beads on a string

Consider N identical beads on a exible string whose endpoints are connected to walls. Each bead has a mass m, and we neglect the mass of the string itself. In the equilibrium position, depicted at the top in Fig. 4.3, the tension in the string is and the beads are separated from each other and the walls by a distance l. The beads are allowed to move in one transverse direction, and we denote the transverse displacement of the p-th bead from the equilibrium position by zp , p = 1, . . . , N , see Fig. (4.3). It turns out to be convenient to introduce the notation z0 and zN +1 for the displacement of the points where the string is connected to the walls. By assumption z0 = zN +1 = 0. (4.15)

When the beads are displaced from equilibrium the segments of string connecting them are no longer horizontal. We call the angle between the horizontal and the segment connecting the p-th and p + 1-st string p and assume in the following that all these angles are small p 1, p = 1, . . . , N. (4.16)

31

...

0 z1

p1 z z2 ...

p
p

... z
p+1

zN

Figure 4.3: N beads on a string We can therefore use the small angle approximation for trigonometric functions sin p tan p p , and cos p 1, p = 1, . . . , N. (4.18) p = 1, . . . , N (4.17)

The total force Fp acting on the p-th bead is the dierence between the tensions in the string segments meeting at that bead. Decomposing into horizontal and vertical components we have from Fig. (4.3)
horizontal Fp = cos p1 + cos p .

(4.19)

For small angles, it follows from (4.18) that


horizontal Fp 0.

(4.20)

For the vertical component of the force we nd, by looking at Fig. (4.3),
vertical Fp = sin p1 + sin p .

(4.21)

32

Now we express the angles p in terms of the vertical displacements via zp zp1 tan p1 = , (4.22) l noting that, because of our convention (4.15), this formula holds for all p = 1, . . . , N . Using the approximation (4.17) we obtain vertical (4.23) Fp = (zp zp1 ) + (zp+1 zp ). l l Thus, Newtons second law yields the equations of motion mz p = (zp zp1 ) (zp zp+1 ), (4.24) l l where p runs from 1 to N . Because of the condition (4.15) the equation for the rst and N -th beads are mz 1 = z1 (z1 z2 ) l l z N = (zN zN 1 ) zN . (4.25) l l In the special case N = 2 of two coupled beads we obtain the coupled equations z 1 = z1 (z1 z2 ) ml ml z 2 = (z2 z1 ) z2 . (4.26) ml ml With the replacements z1 x1 , z2 x2 and /(lm) K/m = k/m, these are the same equations (4.1) that we found for the coupled springs in the previous section when, in the notation of Fig 4.1 k = K . Applying our experience gained there we can decouple the equations by introducing the normal mode coordinates y1 = z1 + z2 , y2 = z1 z2 (4.27)

and the normal mode angular frequencies 1 and 2 3 2 2 1 = , 2 = . ml ml The equations (4.26) are then equivalent to
2 y 1 = 1 y1 2 y 2 = 2 y2

(4.28)

(4.29)

and solved by y1 (t) = A1 cos(1 t) + B1 sin(1 t) y2 (t) = A2 cos(2 t) + B2 sin(2 t) where A1 , B1 , A2 , B2 are arbitrary real constants. We can visualise the normal modes as follows. When the beads move in the normal mode y1 (i.e. y2 = 0) the beads oscillate together. When y1 = 0 and the beads move in the normal mode y2 , they move in opposition: when one is up the other is down and vice-versa. 33 (4.30)

4.3

The eigenvector method for nding normal modes

In this subsection we introduce a systematic way of nding normal modes which also works for N > 2 beads. To motivate it, we re-write the results of the previous section in matrix form. The inverse of the relation (4.27) is 1 z1 = (y1 + y2 ), 2 1 z2 = (y1 y2 ). 2 (4.31)

Hence, when the beads move in the normal mode y1 given in (4.30) with y2 = 0, the displacements z1 and z2 at time t can be combined into the two component vector z1 (t) z2 (t) 1 1 = (A1 cos(1 t) + B1 sin(1 t)) . 1 2 (4.32)

When the beads move in the normal mode y2 , with y1 = 0, their displacements z1 and z2 at time t are given by z1 (t) z2 (t) 1 1 . = (A2 cos(2 t) + B2 sin(2 t)) 1 2 (4.33)

The key property of the vectors (4.32) and (4.33) is that in both cases all components oscillate with the same frequency. The following method, called the eigenvector method, takes this as the starting point for nding normal modes. We illustrate how it works using the example for two beads on a string. Step 1: Write the equations of motion (4.26) in matrix form: z 1 z 2
2 = 0

2 1 1 2

z1 , z2

(4.34)

2 where we introduced the abbreviation 0 = /(ml) (Check that this really is the same equation!)

Step 2: Try to nd a non-zero solution of the form z1 (t) z2 (t) = (A cos(t) + B sin(t)) v1 , v2 (4.35)

where and A, B, v1 , v2 are unknown real constants. Inserting into (4.34) leads to 2 ((A cos(t) + B sin(t)) v1 v2
2 = 0 ((A cos(t) + B sin(t))

2 1 1 2

v1 (4.36) . v2

Assuming that A and B are not both zero, this can only be true for all t if 2 i.e. if 2 1 1 2 v1 v2 2 = 2 0 34 v1 . v2 (4.38) v1 v2
2 = 0

2 1 1 2

v1 , v2

(4.37)

Step 3: Solve the matrix equation (4.38). This is an eigenvalue equation for the matrix M= 2 1 . 1 2 (4.39) and vectors v

We need to nd its eigenvectors and eigenvalues, i.e. the real numbers such that M v = v.

(4.40)

In this case it is straightforward to solve the eigenvalue problem - see example sheets. The matrix M has the eigenvalue 1 = 1 with eigenvector v1 = and eigenvalue
2

1 1

(4.41)

= 3 with eigenvector v2 = 1 . 1 (4.42)

Step 4: Obtain a solution of the form (4.35) for every eigenvalue and eigenvector found in Step 3. Comparing (4.38) and (4.40), we have = Hence
1

2 . 2 0

(4.43)

= 1 is equivalent to 1 = 0 , (4.44)

and inserting the eigenvector (4.41) into (4.35) we obtain the solution z1 (t) z2 (t) = (A cos(1 t) + B sin(1 t)) 1 . 1 (4.45)

Renaming the arbitrary constants A and B as A1 /2 and B1 /2 we recover the solution (4.32). Similarly, from 2 = 3 we deduce 2 = 30 . (4.46) Inserting the eigenvector (4.42) into (4.35) we obtain the solution z1 (t) z2 (t) = (A cos(2 t) + B sin(2 t)) 1 . 1 (4.47)

Renaming the arbitrary constants A and B as A2 /2 and B2 /2 we recover the solution (4.33).

35

We end this section by considering the case of an arbitrary number of beads. According to the eigenvector method, step 1 is to write the equations of motion (4.24) as a matrix equation: z1 z 1 2 1 0 0 0 z 1 2 1 0 0 z2 2 . . 2 . 0 1 2 1 0 (4.48) = , . . . 0 ... 0 0 0 1 2 zN z N
2 where we introduced the abbreviation 0 = /(ml). Next, we implement step 2 by looking for a solution of the form z 1 (t) v1 z 2 (t) v2 . . (4.49) . = (A cos(t) + B sin(t)) . . . z N (t) vN .

Inserting this into (4.48) leads to the eigenvalue problem v1 2 1 0 0 0 1 2 1 0 0 v2 2 . 0 1 2 1 0 . = 2 . 0 ... 0 0 0 1 2 vN

v1 v2 . . . . vN

(4.50)

Solving this is step 3 in our method. We give the solution in the following lemma. The proof is left as an exercise on Sheet 9. Lemma 4.3.1 The matrix 2 1 0 0 1 2 1 0 M = 0 1 2 1 ... 0 0 0 1 has N eigenvalues and eigenvectors given by
n sin N +1 sin 2n N +1 . . vn = . N n sin N +1

0 0 0 2

(4.51)

(4.52)

= 2 2 cos

n N +1

where n = 1, . . . , N . 36

0.8

0.6

0.4

0.2

0
1

0.5

0.5

1 0 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 1 2 3 4

Figure 4.4: Normal modes of three beads on a string Finally, we carry out step 4 by inserting our results for the eigenvalue problem into the solution (4.49). Renaming the constants A and B to An and Bn for the solution corresponding to the n-the eigenvalue n we obtain N independent solutions. With the abbreviation z1 z2 . z= . (4.53) . zN we have the N solutions z n (t) = (An cos(n t) + Bn sin(n t)) v n , where
2 n = 2 n 0 .

(4.54)

(4.55)

37

The eigenvector method produces N solutions, containing two arbitrary real constants An and Bn each. The general solution of (4.34) is a linear combination of these. It contains 2N arbitrary constants as we expect for a general solution of a system of second order dierential equation for N functions. The eigenmodes for N = 3 are shown in Fig. 4.4. Example 4.3.2 Consider the case of N = 3 beads on a string. Each bead has mass m = 10 gram. If the tension in the string is = 1 N and the beads are separated from each other and the wall by l = 1 m, nd the normal modes of the system. .
2 With the data given we nd 0 = /(ml) = 0.01 s2 , so that 0 = 0.1 s 1 . The angular frequencies of the three normal modes are given by (4.55) with n given by (4.52). Inserting N = 3 and n = 1, 2, 3 we nd 2 2 2 2 2 2 1 = (2 2)0 , 2 . (4.56) = 20 , 3 = (2 + 2)0

The normal modes are given by (4.54) with v1 = 1 ,


1 2 1 2

1 v2 = 0 , 1

v 3 = 1 .
1 2

1 2

(4.57)

he displacements of the beads at time t = 0 for each of these normal modes are shown in Fig. (4.4). .

38

Waves

In this section we derive the wave equation for transverse waves on string. We consider a string of length L and with a constant mass density per unit length stretched between two walls. We can think of this system as the N limit of the N beads on a massless string considered in the previous section. This may sound complicated, but we shall see in this section that a string with constant mass density is actually easier to describe mathematically than a nite but large number of beads on a (massless) string. 5.1 The wave equation

Consider the general displacement of the string from equilibrium shown in Fig. (5.1). The string is allowed to move in one transverse direction (the vertical direction in the gure). We introduce a coordinate x [0, L] along the string and denote and the transverse displacement by z . Since the transverses displacement depends on the time t and on x, z is a function of t and x. The fact that the string is xed at its endpoints is mathematically expressed via z (t, 0) = z (t, L) = 0, t R
(t,x+ x) (t,x) z(t,x) x z(t,x+ x)

(5.1)

x+ x

Figure 5.1: The stretched string In order to derive an equation of motion for z , consider a small segment (x, x + x) of the string. The angle between the string and the horizontal is denoted by ; since the angle varies along the string and also with time, is a function of t and x. As in the derivation of the equation of motion for beads on a string we assume that is small, i.e. (t, x) so that sin (t, x) tan (t, x) (t, x), and cos (t, x) 1, t R, x [0, L]. 39 (5.4) t R, x [0, L], (5.3) 1, t R, x [0, L], (5.2)

Next we consider the forces acting on the segment (x, x + x). Denoting the tension in the string by , the horizontal force on the segment is F horizontal (t, x) = cos (t, x) + cos (t, x + x). However, with the approximation (5.4) we nd F horizontal (t, x) 0. For the vertical component of the force we nd F vertical (t, x) = sin (t, x) + sin (t, x + x). Using the approximation (5.3) we simplify the right hand side sin (t, x + x) sin (t, x) (t, x + x) (t, x). Furthermore, the Taylor expansion (t, x + x) = (t, x) + x implies, for suciently small x, (t, x + x) (t, x) Hence the vertical component of the force is F vertical (t, x) (t, x)x. x (5.11) (t, x)x. x (5.10) (t, x) x + terms of order(x)2 (5.9) (5.8) (5.7) (5.6) (5.5)

Now we express the angles (t, x) in terms of the vertical displacements via tan (t, x) = z (t, x + x) z (t, x) z , x x (5.12)

where we have used the Taylor expansion again, and assumed that x is small. For small angles we therefore have (t, x) so that F vertical (t, x) 2z (t, x)x. x2 (5.14) z , x (5.13)

In order to derive the equations of motion for the string we apply Newtons second law to the string segment [x, x + x]. Recalling that the string has constant mass per unit length , we note that segment [x, x + x] has mass x. We approximate the transverse 40

displacement of that the segment by z (t, x), which becomes exact in the limit x 0. 2z and we obtain the equation of motion Hence the acceleration of the segement is t2 x 2z 2z = x. t2 x2 (5.15)

Dividing by x we have the partial dierential equation 1 2z 2z = , v 2 t2 x2 (5.16)

where the parameter v is dened in terms of the string tension and the mass density v= . (5.17)

The equation (5.16) is called the wave equation. It is one of the most important equations of mathematical physics, arising in numerous applications of mathematics to physics. Before we describe a systematic approach to solving the wave equation in the next section, we consider a simple solution. It is easy to check that the function z (t, x) = cos(t) sin(kx) solves the wave equation (5.16) for any pair of real parameters and k satisfying = vk. (5.19) (5.18)

The solution (5.18) is periodic in both space and time. We have already discussed the periodicity of cos(t) in our treatment of simple harmonic motion. Since cos( (t + T )) = cos(t) we have the familiar formula for the period T = 2 (5.21) (5.20)

However, the solution (5.18) is also periodic in space. The equation sin(k (x + )) = sin(kx) holds for all x if we take = 2 . k (5.23) (5.22)

The parameter characterises the spatial periodicity of the wave; it is called the wavelength.

41

5.2

Standing waves

The solution (5.18) does not, in general, satisfy the boundary condition (5.1) at x = L. When we require z (t, L) = 0 t R we get the extra condition sin(kL) = 0. This imposes a condition on the allowed values for k . The equation (5.25) holds if k= n , L n = 1, 2, 3 . . . . (5.26) (5.25) (5.24)

We thus obtain n solutions of the wave equation (5.16) satisfying the boundary condition (5.24): zn (t, x) = cos(n t) sin( where n = v n . L (5.28) n x), L (5.27)

In this section we explain a systematic way of obtaining solutions like (5.27). The method is called separation of variables and is useful in solving many partial dierential equations. The idea is to solve the partial dierential equation (5.16) by assuming that the unkown function z (t, x) is a product of an unknown function g (t) which depends only on t and an unknown function f (x) which depends only on x. Thus we insert z (t, x) = g (t)f (x) (5.29)

into (5.16) and hope that we can solve the resulting equations for g and f . It follows from (5.29) that 2z d2 g ( t, x ) = (t)f (x) t2 dt2 and 2z d2 f ( t, x ) = g ( t ) (x). x2 dx2 The insertion of (5.29) into the wave equation leads to 1 d2 g d2 f ( t ) f ( x ) = g ( t ) (x). v 2 dt2 dx2 Assuming that neither f nor g vanish this is equivalent to 1 d2 g 1 d2 f ( t ) = (x). v 2 g (t) dt2 f (x) dx2 42 (5.33) (5.32) (5.31) (5.30)

The left hand side of this equation depends only on t, and the right hand side depends only on x. The only way two functions of dierent variables can be identical is if they are both constants. Assuming for now that the constant is negative, we denote it by k 2 . Thus we obtain two ordinary dierential equations, one for the function f which depends on x d2 f = k 2 f dx2 and one for the function g which depends on t only: d2 g = v 2 k 2 g. dt2 (5.35) (5.34)

Our assumption (5.29) has thus allowed us to trade one partial dierential equation for two ordinary dierential equations with an arbitrary parameter k . Moreover, both equations are the familiar dierential equation for simple harmonic motion. Thus the general solution of (5.34) is f (x) = cos(kx) + sin(kx) with , arbitrary real constants. The condition (5.1) implies f (0) = 0, f (L) = 0. (5.37) (5.36)

In order to satisfy f (0) = 0 we must have = 0. The requirement f (L) = 0 imposes a condition on k , which we already discussed after (5.25). Thus we conclude that k must take one of the values k= n , L n = 1, 2, 3 . . . . (5.38)

Inserting these values into (5.35) we obtain d2 g n = v 2 2 dt L


2

g.

(5.39)

For each value of n, this is again the equation for simple harmonic motion. So, for each value of n, we get the general solution gn (t) = An cos(n t) + Bn sin(n t) where n = v n . L (5.41) (5.40)

Finally assembling f and gn into a function of t and x according to (5.29) we obtain the following solutions of the wave equation and the boundary condition (5.1): zn (t, x) = (An cos(n t) + Bn sin(n t)) sin 43 n x . L (5.42)

Here we have set the constant = 1 in (5.36). This can be done without loss of generality because multiplies the the constants An and Bn which are themselves arbitrary. In analogy to the solutions (4.54) the solutions (5.42) are called the the normal modes of the string. The mode n = 1 z1 (t, x) = (An cos(1 t) + Bn sin(1 t)) sin x . L (5.43)

is called the fundamental mode of the string. The solutions for n > 1 are called higher modes or, in accoustics, overtones. Each mode has a characteristic angular frequency, period and wavelength. With the angular frequency of the n-th mode given by (5.41), the period of the n-th mode is Tn = 2 2L = . n vn (5.44)

The wavelength can be computed from (5.38) via (5.23). We nd n = 2


n L

2L . n

(5.45)

The following example illustrates the importance of normal modes. Example 5.2.1 Consider a string of length L = 1m, with uniform mass density and total mass M = 0.1 kg. The string is stretched so that the tension in the string is = 10 Newton, and xed at its endpoints. It is allowed to oscillate freely in one transverse direction. (a) Find all normal modes of the string and give their angular frequencies, periods and wavelengths. Sketch the lowest three modes at time t = 0. (b) If the string is given an initial displacement z0 (x) = sin(x) and released from rest, nd its subsequent motion. (a) The mass density is related to the total mass and the length of the string via = M/L. Hence, with the values given in the example, = 0.1kg/m. Thus v= = 100m s1 = 10m s1 .

Inserting the numerical values for v and L into (5.42) we nd zn (t, x) = (An cos(10nt) + Bn sin(10nt)) sin(nx). (5.46)

The angular frequency of the n-th mode is n = 10n s1 , the period is Tn = 51 s and the n 2 wavelength is n = n m. The functions zn (0, x) are sketchted for n = 1, 2, 3 in Fig 5.2.

44

1 0.8 0.6 0.4 0.2 z 0 0.2 0.4 0.6 0.8 1 1 0.8 0.6 0.4 0.2 z 0 0.2 0.4 0.6 0.8 1 1 0.8 0.6 0.4 0.2 z 0 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

Figure 5.2: Lowest normal modes of a string of length L = 1 m (n = 1, 2, 3) (b) The information is released from rest means that the required solution should have vanishing time derivative at time t = 0. Thus we seek a solution of the wave equation and the boundary condition (5.1) which also satises the intial conditions zn (0, x) = z0 (x), z (0, x) = 0 t (5.47)

We check if one of the normal modes (5.46) satises these conditions. We have zn (0, x) = An sin(nx) and zn (0, x) = 10nBn sin(nx) t (5.49) (5.48)

Hence we can satisfy (5.47) by taking n = 1, A1 = 1 and B1 = 0, and the required solution is z (t, x) = cos(10t) sin(x). (5.50)

45

In the example we were lucky that we could satisfy the initial conditions with one of the normal modes of the string. For more general initial conditions this will not be possible. The next section introduces a tool for tackling general intial value problems for the string. The tool is called Fourier analysis, and widely used in applied mathematics. 5.3 Fourier series

We consider functions on the interval [L, L], where L > 0, and show that any reasonable function f : [L, L] R can be respresented as a series nx nx an cos + bn sin f (x) = a0 + L L n=1 n=1

(5.51)

with a0 , an and bn , n = 1, 2, 3 . . .. We are going to give an explicit formula for these coecients further below. To motivate the formula, we recall the denition of an inner product space, and the expansion of a vector in such a space in an orthognal basis. Example 5.3.1 The vector space R2 has the inner product y x1 , 1 y2 x2 (a) Check that the vectors v 1 = = x1 y1 + x2 y2 . 1 1 (5.52)

1 and v 2 = 1 with respect to the inner product , . 1 2

form an orthogonal basis of R2

(b) Expand the vector v =

in the basis {v 1 , v 2 }

(a) Using the formula (5.52) we nd v 1 , v 2 = 1 1 = 0 so that v 1 and v 2 are indeed orthogonal. They form a basis since any two orthogonal vectors form a basis of the two-dimensional vector space R2 . (b) We are looking for coecients c1 and c2 such that v = c1 v 2 + c2 v 2 . In order to compute the coecient c1 , we take the inner product with v 1 to nd v 1 , v = c1 v 1 , v 1 and hence 3 3 = 2c1 c1 = . 2 Similarly, to nd the coecient c2 we take the inner product with v 2 to nd v 2 , v = c1 v 2 , v 2 46 (5.56) (5.55) (5.54) (5.53)

and hence 1 1 = 2c2 c2 = . 2 (5.57)

The formulae (5.54) and (5.56) can be summarised in one formla as cn = vn, v vn, vn n = 1, 2. (5.58)

In this formulation it holds for any vector space with inner product. In order to apply it to (5.51) we need to consider the space of all functions on [L, L] as a vector space, and equip it with an inner product. More precisely, let H = space of all piecewise continuous functions on[L, L] and dene the inner product of two functions f, g H as
L

(5.59)

f, g =
L

f (x)g (x)dx.

(5.60)

Having dened the vector space H and the inner product we want to expand an arbitray element of H in a xed orthogonal set. The set we want to use is S = {1, cos Thus we need to show Lemma 5.3.2 The set S given in (5.61) is an orthogonal set in H . Proof : The following general results will be useful in the proof. Recall that a function F : [L, L] R is called odd if F (x) = F (x) and even if F (x) = F (x). 1. The integral of any odd function F : [L, L] R from L to L vanishes. i.e.
L

2x x 2x x , cos , . . . , sin , sin , . . .}. L L L L

(5.61)

F (x) = F (x)
L

F (x)dx = 0.

(5.62)

2. For a non-zero integer N we have


L

cos
L

N x dx = 0 L

(5.63)

The proof of the rst result is an exercise on the example sheets, and the proof of the second result is an elementary integration. The results allow us to deduce immediately that the function 1 is orthogonal to all other elements of S : nx 1, sin = L
L

sin
L

nx =0 L

(5.64)

47

since the sine function is odd and, by (5.63), 1, cos Next consider the inner products cos mx nx , sin = L L
L

nx = L

cos
L

nx = 0. L

(5.65)

cos
L

nx mx sin dx. L L

(5.66)

Now use the trigonometric identity 2 sin A cos B = sin(A + B ) + sin(A B ) and the fact that the sine function is odd to deduce cos nx 1 mx , sin = L L 2
L

sin
L

(m + n)x (n m)x + sin dx = 0. L L

(5.67)

Using 2 cos A cos B = cos(A + B ) + cos(A B ) we nd cos mx nx 1 , cos = L L 2


L

cos
L

(m + n)x (n m)x + cos dx. L L

(5.68)

Integrating the rst terms gives zero by (5.63), and the same holds for the second term except when n = m. Thus cos mx nx , cos = L L 0 if n = m L if n = m. (5.69)

Finally, using 2 sin A sin B = cos(A B ) cos(A + B ) we nd sin nx 1 mx , sin = L L 2


L

cos
L

(m n)x (n + m)x cos dx. L L

(5.70)

This time integrating the second terms gives zero by (5.63), but the rst term only gives zero when n = m. Thus sin mx nx , sin = L L 0 if n = m L if n = m (5.71)

Hence the set S is an orthogonal set. Moreover we note the inner products of the elements with themselves 1, 1 = 2L, sin nx nx nx nx , sin = cos , cos =L L L L L (5.72) . The expansion nx nx f (x) = a0 + an cos + bn sin L L n=1 n=1 48

(5.73)

of an element f H is therefore an expansion in an orthogonal set. Hence we compute the coecients a0 , an and bn in analogy to (5.58) via a0 = an = bn =
L 1, f 1 = f (x)dx 1, 1 2L L ,f cos nx 1 L nx L = f (x) cos dx nx nx cos L , cos L L L L sin nx ,f 1 L nx L f (x) sin dx nx = nx sin L , sin L L L L

(5.74)

The series (5.73) with the coecients dened by (5.74) is called the Fourier series of f. Example 5.3.3 Find the Fourier series for f (x) = x, x [1, 1] With L = 1 the formulae (5.74) give a0 = an = bn 1 2
1 1

xdx = 0 since the integrand is odd


1

nx dx = 0 since the integrand is odd L 1 1 nx 1 1 = f (x) sin dx = [x cos(nx)]1 1 + L n n 1 2 = (1)n+1 n x cos

cos(nx)dx =
1

(5.75)

Hence f (x) 2 1 1 [sin(x) sin(2x) + sin(3x) . . . 2 3 (5.76) . Even though the functions we consider are only dened on the interval [L, L], the Fourier which is series makes sense for all x R. It converges almost everywhere to a function f a periodic extension of f i.e. (x) = f (x 2nL) for (2n 1)L < x < (2n + 1)L, f nZ (5.77)

is not continuous, the Fourier series converges to the At points x where the function f at x.1 This phenomenon is called the average of the right-limit and the left-limit of f Gibbs phenomenon.
The right-limit of f at x is dened as lim ), with > 0.
1 0

f (x+ ), with > 0 Similarly, the left-limit is lim

f (x

49

5.4

Using Fourier series for solving the wave equation

In our study of standing waves on a string, the dependence of displacement z on x was given by a function f on the interval [0, L]. We now show how to obtain a series for such functions which involves only sine terms (or only cosine terms). The resulting series is called the half range Fourier series. We start with the Fourier sine series. Given the function f : [0, L] R extend it to an odd function f : [L, L] R by dening f (x) = f (x). Then the new function f : [L, L] R has a (full range) Fourier series

f (x) = a0 +
n=1

an cos

nx nx + bn sin L L n=1

(5.78)

with coecients dened as in (5.74). However, since f is odd we can use the rule (5.62) are odd, their to deduce that a0 = an = 0 in (5.78). Moreover, since both f and sin nx L product is even and therefore bn = 1 L
L

f (x) sin
L

nx 2 dx = L L

f (x) sin
0

nx dx L

(5.79)

Thus bn is dened purely in terms of the original function f : [0, L] R . The series

f (x) =
n=1

bn sin

nx L

(5.80)

with bn dened via (5.79) is called the Fourier sine series of f : [0, L] R The Fourier cosine series of a function f : [0, L] R is obtained by extending it to an even function f : [L, L] R by dening f (x) = f (x). Then the new function f : [L, L] R has a full range Fourier series nx nx f (x) = a0 + an cos + bn sin L L n=1 n=1

(5.81)

with coecients dened as in (5.74). Since f is even we can now use the rule (5.62) to are even, their product deduce that bn = 0 in (5.81). Moreover, since both f and cos nx L is even, too, and therefore a0 = = an 1 2L 1 = = L 1 L f (x)dx L 0 L L nx 2 L nx f (x) cos dx = f (x) cos dx L L 0 L L f (x)dx =
L

(5.82)

Thus a0 , an are dened purely in terms of the original function f : [0, L] R . The series

f (x) = a0 +
n=1

an cos

nx L

(5.83)

with a0 , an dened via (5.82) is called the Fourier cosine series of f : [0, L] R 50

Example 5.4.1 Find the Fourier sine series of the function f (x) = x2 , x [0, ] From (5.79) we nd, with L = , bn = 2
L

x2 sin(nx)dx
0

(5.84)

Integrating by parts twice and using sin(n ) = 0 and cos(n ) = (1)n we nd 2 4 4 bn = (1)n + 3 (1)n 3 . n n n

(5.85)

Finally, we illustrate the application of Fourier series to initial values problems for the string. Example 5.4.2 0 < x < . (a) Find the Fourier sine series of the function f (x) = x( x), with

(b) Consider a string of length L = meters, with tension = 10 Newton and mass per unit length = 0.1 kg /m. Using the result of (a) nd the solution z (t, x) of the wave equation for the transverse displacement of the string satisfying the initial conidition z z (0, x) = x( x), (0, x) = 0. t (a) The Fourier sine series on the intervl [0, ] takes the form

f (x) =
n=1

bn sin(nx),

(5.86)

with bn = = = = = = so

2 f (x) sin(nx)dx 0 2 x( x) sin(nx)dx 0 1 2 1 x( x) cos(nx) + n n 0 2 1 2 ( 2x) 2 sin(nx) + 2 n n 0 4 1 cos(nx) 2 n n 0 4 (1 (1)n ) n3

( 2x) cos(nx)dx
0

sin(nx)dx
0

(5.87)

x( x) =
n=1

4 (1 (1)n ) sin(nx) n3 sin(x) + 1 1 sin(3x) + sin(5x) . . . . 27 125 (5.88)

51

(b) With the parameters given in the question, we nd v = / = 10m s1 . Hence, using the general formula (5.42), the normal modes of the string are those found already found in (5.46): zn (t, x) = (An cos(10nt) + Bn sin(10nt)) sin (nx) . (5.89)

It is easy to check (we will do this in the next section) that arbitrary linear combination of the normal modes also satisfy the wave equation. In particular (ingoring convergence issues), the innite sum

z (t, x) =
n=1

(An cos(10nt) + Bn sin(10nt)) sin (nx)

(5.90)

solves the wave equation and satises the boundary condition z (t, 0) = z (t, ) = 0. We now need to choose the constants An and Bn so that the given initial conditions are also satised. Since z (t, x) = t

(10An sin(10nt) + 10Bn cos(10nt)) sin (nx)


n=1

(5.91)

we have the initial velocity z (0, x) = t

10Bn sin (nx) .


n=1

(5.92)

To make this vanish, we simply choose Bn = 0 for all n. The initial displacement is

z (0, x) =
n=1

An sin (nx) .

(5.93)

We would like this to be equal to x( x). Comparing with the sine series (5.88), we see that we can achieve this by setting An = 4 (1 (1)n ). n3 (5.94)

Hence the required solution of the initial value problem for the vibrating string is

z (t, x) =
n=1

4 (1 (1)n ) cos(10nt) sin (nx) . n3

(5.95)

52

To end this section, we collect the similarities between eigenvector method for studying the motion of beads on a string and the separation of variables method for solving the motion of a string in a table.

N beads on a string

Vibrating string

equation of motion

d2 z = Mz 2 m dt M vn = = 2 2 cos
n nv

2z 2z = t2 x2 d2 fn 2 = kn fn dx2 kn = n L nx L L nm 2

normal mode equation eigenvalues

n N +1

normal modes

n vj = sin

nj N +1 N +1 nm 2 m
n L

fn (x) = sin

vn, vm = normal frequencies


N 2 n =

fn (x)fm (x)dx =
0 2 n =

2 k n

general solution
n=1

(An cos(n t) + Bn sin(n t)) v n


n=1

(An cos(n t) + Bn sin(n t)) fn (x)

Table 5: The eigenvector method versus the separation of variables

53

5.5

Travelling waves

In this subsection we return to the wave equation 1 2z 2z = , v 2 t2 x2 (5.96)

and discuss a class of solutions called travelling waves. So far we have only considered standing waves of the form zk (t, x) = cos(t) sin(kx). (5.97)

However, using the trigonometric identity 2 cos A sin B = sin(A + B ) + sin(B A) we can write this solution as zk (t, x) = 1 (sin(kx + t) + sin(kx t)) . 2 (5.98)

Now it is easy to check that both zk,+ (t, x) = sin(kx + t) and zk, (t, x) = sin(kx t) satisfy the wave equation (5.96) provided that the relation = vk (5.101) (5.100) (5.99)

is satised. Note that neither of these solutions satisfy the boundary condition z (t, 0) = z (t, L) = 0. However, recall that our derivation of the wave equation in Sect. 5.1 was independent of the boundary condition. Hence we can think of the wave equation without boundary conditions as a mathematical description of the transverse displacement of an innitely long string. This is the point of view we take in this section: the range of x is all of R, and no boundary conditions are imposed. In particular, we therefore do not have any conditions on the parameter k : it is an arbitrary real number. We may assume without loss of generality that k > 0 since changing the sign of k in (5.99) and (5.100) does not lead to an essentially new solution. Similarly, we assume without loss of generality that > 0. Both the solution (5.99) and (5.100) are examples of travelling waves. To understand in what sense they travel consider the spatial dependence of, for example, z+ at time time t = 0: zk,+ (0, x) = sin(kx) A time interval t > 0 later the wave is zk,+ (t, x) = sin(kx + t). 54 (5.103) (5.102)

and which vanishes at the origin, the While (5.102) is a sine function with period T = 2 function (5.103) (viewed as a function of x) , is a sine function with the same period but shifted to the left by x = t. Using (5.101) we have k

x = v t.

(5.104)

Since this holds for an arbitrary interval t we can visualise the solution z+ as a wave which moves to the left with speed v , see Fig. (5.3).
1

0.5

0.5

Figure 5.3: Left moving wave: plot of sin(2x + t) for t = 0 (solid line) and t = 1 (dashed line) A similar analysis for z leads to the conclusion that it is a wave which moves to the right with speed v , see Fig. (5.4).
1

0.5

0.5

Figure 5.4: Right moving wave: plot of sin(2x t) for t = 0 (solid line) and t = 1 (dashed line) We thus arrive at the remarkable conclusion that the standing wave (5.97) is a sum of a left- and a right-moving wave. Moreover we found the physical interpretation of the parameter v : it is the speed of travelling wave solutions of the wave equation (5.96). It is surprisingly easy to write down innitely many travelling wave solutions, as shown by the following theorem. Theorem 5.5.1 Let f : R R be an arbitrary twice dierentiable function. Then z+ (t, x) = f (x + vt) and z (t, x) = f (x vt) both satisfy the wave equation (5.96). 55

Proof: Let u = x + vt. Then df u df z+ = =v t du t du and


2 2 z+ 2d f =v . t2 du2

Also

z+ df u df = = x du x du d2 f 2 z+ = . x2 du2

and

1 2 z+ d2 f 2 z+ = = . v 2 t2 du2 x2 Similarly dening w = x vt, we have z df w df = = v t dw t dw and


2 2 z 2d f = v . t2 dw2

Hence

Also

z df w df = = x dw x dw 2 z d2 f = . x2 dw2 1 2 z d2 f 2 z = = . v 2 t2 dw2 x2

and

Hence

Using this theorem, we can construct waves of arbitrary shape. Taking for example the Gaussian f (u) = exp(u2 ) we can construct a right travelling wave z (t, x) = exp((x vt)2 ) which, at any time t, looks like a bell centered at vt, see Fig 5.5. We have already seen how to construct a standing wave by adding two travelling waves, one travelling to the right and one travelling to the left. In fact we can add any two solutions of the wave equation to obtain another solution. Although this result is easy to prove (see below), it is of fundamental importance in the study of the wave equation. It is called the superposition principle. We state in the following form: Theorem 5.5.2 Suppose z1 (t, x) and z2 (t, x) satisfy the wave equation (5.96) and A, B are real constants. Then z (t, x) = Az1 (t, x) + Bz2 (t, x) also satises the wave equation. 56

0.8

0.6

0.4

0.2

Figure 5.5: Right moving bell shape: plot of exp((x t)2 ) for t = 0 (solid line) and t = 1 (dashed line) Proof: The result follows from the linearity of the wave equation. With z (t, x) as dened in the theorem we deduce z z1 z2 =A +B t t t and 2 z1 2 z2 2z =A 2 +B 2 . t2 t t Similarly z z1 z2 =A +B x x x and 2z 2 z1 2 z2 = A + B . x2 x2 x2 Hence 1 2z 1 2 z1 1 2 z2 = A + B v 2 t2 v 2 t2 v 2 t2 2 2 z1 z2 = A 2 +B 2 x x 2 z = , x2

(5.105)

where we used that both z1 and z2 satisfy the wave equation in the intermediate step.

57